Research On Compilation-Based Error Mitigation Technologies For Quantum Computing Programs

Quantum computing is a new computing paradigm that promises to deliver computing power far beyond that of classical computing for certain problems.With the recent rapid development of technologies,quantum computing has entered the noisy intermediate-scale quantum era.However,quantum computers in this era have not yet achieved quantum error correction.The noise in the hardware constantly affects the computational process and leads to frequent computational errors,which seriously hinders the practicalization of quantum computing.Therefore,research into error mitigation technologies is urgently needed.In the current era,limited quantum error perception capabilities and limited quantum control methods have greatly increased the difficulty of designing effective error mitigation techniques.Crosstalk,systematic noise,and the intertwined effects of multiple errors have become the three key issues that need to be urgently addressed by error mitigation.This thesis stands at the system level and conducts in-depth research on compilation-based error mitigation technologies,with a focus on solving the above challenges and problems.The main contributions of this thesis are:(1)Aiming at crosstalk,this thesis proposes a quantum gate reordering technique.This technique uses commutativity rules to rearrange the execution order of quantum gates,significantly reducing unwanted parallel patterns that can cause severe crosstalk.This technique shows its value at both theoretical and practical levels.Theoretically,it proposes a complete set of generalized commutativity rules and a reordering scheme evaluation criterion that takes into account various errors.And practically,it proposes a reordering algorithm that can efficiently find good solutions in polynomial time.Evaluations on 117 programs show that the probability of successful trials of these programs can be increased by up to a factor of 2.2×.(2)Targeting systematic noise,this thesis proposes a pulse-scheduling cooptimization technique to suppress the systematic ZZ noise in superconducting quantum computers.This technique realizes suppression purely in a software manner,overcoming the problem of relying on special hardware structures in the existing methods.To address the challenge of exponential complexity in pulse optimization,this thesis co-designs pulse optimization objectives,identity gate insertion strategies,and quantum gate layering strategies.These successfully control the complexity of the overall scheme to the polynomial level.Simulations on 4 ～ 12 qubits show that this technique can improve the fidelity of(the final state of)a program by a factor of 11× on average,and experiments on a real device also demonstrate that it can reduce effective noise strength by about two orders of magnitude.(3)For the intertwined effects of multiple errors,this thesis proposes an adaptive learning method that can find effective error mitigation schemes automatically.This method is based on the framework of probabilistic error cancellation,inheriting its theoretical advantage of being able to mitigate multiple errors simultaneously.And this method uses a learning strategy based on heredity and mutation,which avoids the problem that probabilistic error cancellation requires an explicit characterization of errors,and also reduces the cost of use.Evaluations show that this method can achieve better error mitigation at a lower cost than existing methods.The cost is reduced by about one order of magnitude.(4)Based on the above techniques,this thesis designs and implements a quantum compilation framework for error mitigation technologies.The framework integrates the needs of expressing quantum programs and implementing error mitigation techniques,incorporates all the error mitigation techniques proposed in this thesis,and provides support for future technology development.